- Built fraudulent click classification models (LightGBM) in Kaggle’s TalkingData AdTracking Fraud Detection Challenge.
- Trained models on Amazon Web Services with over 80 million data points in 45 dimensions (derived from 6).
- Built default rate prediction models (XGBoost) from LendingClub’s historical data retrieved via API.
- Cleansed data and engineered new features (variable names consistency check, categorization, string vectorization, etc.)
albert-z-guo / kaggle-competition-solutions-and-data-science-projects Goto Github PK
View Code? Open in Web Editor NEWKaggle’s TalkingData AdTracking Fraud Detection Competition Solutions and LendingClub Data Science Project